Abnormal event detection method and device

Abstract

The embodiment of the invention provides an abnormal event detection method and device. The method comprises the following steps: obtaining POI data and communication data of a current base station within first preset time are acquired, fusing the POI data of the current base station in the first preset time with the communication data; obtaining a current total feature vector of the current basestation; further, using a trained multi-classification model is to calculate and obtain a current category probability vector of the current total feature vector; calculating the Euclidean distance between the current category probability vector of the current base station and the historical category probability vector of the current base station; if the Euclidean distance exceeds a distance threshold value, determining the current base station as an undetermined abnormal base station, juding whether the first difference value between the current average feature vector of the undetermined abnormal base station and the historical average feature vector of the undetermined abnormal base station exceeds a preset first difference threshold value or not, determining whether the current communication data is abnormal or not. The accuracy of abnormal event detection can be improved.

Description

technical field [0001] The invention relates to the technical field of detection, in particular to an abnormal event detection method and device. Background technique [0002] When an unexpected event occurs in a city, such as a large-scale event or a traffic accident, it will cause road crowds or traffic jams. These emergencies are called abnormal events. Detection of abnormal events can effectively guide urban road traffic . [0003] Existing schemes for detecting abnormal events include: detection schemes based on granularity clustering and detection schemes based on density; [0004] In the detection scheme based on granularity clustering, an urban area is firstly divided into multiple areas according to the coverage of communication base stations, and each area has a communication base station responsible for the communication in this area. Classify the communication data in each communication base station according to two granularities, then cluster the communication...

AI Technical Summary

Problems solved by technology

[0005] The above detection scheme based on granularity clustering will have the following problems when implemented: when the communication data is affected by noise, the clustering results of the same communication data at different granularities will be different, and it is easy to misidentify abnormal events

[0007] The above-mentioned density-based detection scheme will have the following problems when it is implemented: when LOF scores a single particle, only a limited number of particles that are close to the single particle are considered, that is, the correlation of a single particle between local particles, when the single particle When the outlier degree of the limited number of particles considered by the particle is large, the LOF of the single particle will also be large, resulting in the particle's score value being affected by local particles, so the accuracy of particle scoring is not high, resulting in accurate identification of abnormal events degree is not high

Images